Complex conjugate

[tlienart]

In cov(x, y, Simple) the docstring specifies * is the complex conjugate. I think the only place where we allow complex would then be through the cov from Statistics.

I wonder if it's not more consistent (at least for now) to just say that we only consider real data matrices throughout the package and then extend (much) later on if there's a use case?

[mateuszbaran]

I'm actually thinking about extending Simple covariance to all cases covered by cov from StatsBase.jl. It would be nice to have a common interface for all methods. All other algorithms we have should definitely be clearly marked as designed for real matrices only.